Generative Adversarial Networks in one picture

Source: Deep Learning on Medium

Generative Adversarial Networks in one picture

Nothing more to say

Generative Adversarial Network (full size version here).

Some years ago I had my first experience with GANs (Generative Adversarial Networks). I took an online course, the video lessons concernig GANs were taught by Ian Goodfellow. I was programming some little snippets for a test-project using CelebA dataset. I admit the smart intuition behind GANs, but I was able to generate only freaks, just creepy monsters… it was partially working in my case (LOL). See the results below…

Some useful links: original paper by Goodfellow et al., Deep Convolutional GANs using TensorFlow (you can run the code in Colab), DCGANs using Pytorch (available for Colab), some tricks for better performance here.